Prerequisites: This page requires Python familiarity. If you are a non-technical operator, work with your developer to configure these classes.
Core class
Conversation
The primary class for managing conversation state at runtime. Every function receives a Conversation instance as its first argument.
def get_conversation_id(conv: Conversation):
return conv.id
For full attributes and methods, see the
conv object page.
Voice classes
Use voice classes to configure TTS providers programmatically — for example, in a start function or when using multi-voice.
VoiceWeighting
Assign specific weightings to voices for multi-voice setups, adjusting their prominence in a given context.
VoiceWeighting(
voice=ElevenLabsVoice(
provider_voice_id="LcfcDJNUP1GQjkzn1xUU",
similarity_boost=0.2,
stability=0.4
),
weight=0.25
)
TTS provider classes
ElevenLabs
Cartesia
PlayHT
Rime
Minimax
Hume
Custom
Configures voice settings for ElevenLabs TTS, with control over stability and similarity.elevenlabs_voice = ElevenLabsVoice(
provider_voice_id="a1b2C3d4E5f6G7h8I9j0",
stability=0.5,
similarity_boost=0.7
)
| Parameter | Description |
|---|
provider_voice_id | The ElevenLabs voice ID |
stability | Consistency of tone and delivery (0.0–1.0) |
similarity_boost | How closely the voice matches the original (0.0–1.0) |
Configures voice settings for Cartesia TTS. Uses Emotion objects for expressive control.from polyai.voice import CartesiaVoice, Emotion, EmotionKind, EmotionIntensity
cartesia_voice = CartesiaVoice(
provider_voice_id="a1b2c3d4",
speed=0.0,
emotions=[Emotion(EmotionKind.POSITIVITY, EmotionIntensity.HIGH)]
)
| Parameter | Description |
|---|
provider_voice_id | The Cartesia voice to use |
speed | Speech rate: -1.0 (slowest) to 1.0 (fastest) |
emotions | List of Emotion objects (see voice reference) |
model_id | "sonic" or "sonic-preview" |
Configures voice settings for PlayHT TTS.playht_voice = PlayHTVoice(
provider_voice_id="en_us_male_1",
style="conversational"
)
| Parameter | Description |
|---|
provider_voice_id | The PlayHT voice to use |
style | Speaking style or tone |
Configures voice settings for Rime TTS.rime_voice = RimeVoice(
provider_voice_id="voice_id",
speech_alpha=1.0,
model_id="mistv2"
)
| Parameter | Description |
|---|
provider_voice_id | Your configured Rime voice identifier. Contact your PolyAI rep or Rime for the correct value. |
speech_alpha | Speech rate multiplier: <1.0 faster, >1.0 slower |
model_id | "mistv2" or "mist" |
Configures voice settings for Minimax TTS.minimax_voice = MinimaxVoice(
model_id="speech-02-hd",
voice_id="voice_id",
speed=1.0,
emotion="happy"
)
| Parameter | Description |
|---|
model_id | "speech-02-hd", "speech-02-turbo", "speech-01-hd", or "speech-01-turbo" |
voice_id | The Minimax voice ID |
speed | Speech rate (0.5–2.0) |
emotion | "happy", "sad", "angry", "fearful", "disgusted", "surprised", "neutral" |
Configures voice settings for Hume TTS.hume_voice = HumeVoice(
provider_voice_id="voice_uuid_or_name",
version="2"
)
| Parameter | Description |
|---|
provider_voice_id | Voice UUID or name |
voice_description | Optional description for voice personality |
version | "1" (octave-1) or "2" (octave-2) |
instant_mode | Ultra-low latency mode (boolean) |
Define voice configurations for a custom TTS provider.voice_config = CustomVoice(
provider="MY_PROVIDER",
provider_voice_id="voice_id"
)
| Parameter | Description |
|---|
provider | Custom TTS provider name |
provider_voice_id | Voice ID within the provider |
For full voice configuration examples including ElevenLabs model IDs, Cartesia emotions, and cache behavior, see the
Voice class reference.
